元資料

隨著電腦與網路科技的運用，網路上有越來越多的資訊，儼然形成了另一個形式的圖書館。如何有效的組織檢索網路上各種不同的資源，一直是資訊使用者與整理者所希望解決的議題。機讀編目格式一直是圖書館作業自動化後整理組織圖書館資源的依據，但是否適合處理網路資源，討論聲音一直不斷。本文擬就機讀編目格式與Dublin Core發展背景，資料欄位架構，並就兩者資料結構來探討不同格式處理網路資源的議題與適用性。

With the growth of the information on the Internet, Internet forms a different type of libraries. It has been an issue on how to organize the information on the Internet and access effectively. The Machine Readable Catalog (MARC) format has been the standard of creating the records for the library collection. Is MARC right for the Internet sources? This article looks into the characters of MARC and Dublin Core to outlines the issues on organizing Internet sources.

資訊檢索研究的目的在解決人類對於資訊的需求。然而，隨著不同資訊型態的出現，資訊檢索的技術亦逐漸多樣化，以適應各種型態的資訊。本文說明透過元資料進行檢索，可以適用於各種型態的資訊，故可稱之為資訊檢索核心技術。作者並探討三種不同層次的元資料：靜態權威元資料、動態權威元資料、個人化元資料，及其可能的應用方式。

The purpose of researches for information retrieval is to fulfill the information need. The various techniques for information retrieval are proposed to adapt to the diversified information types. This paper manages to deliberate the importance of metadata and suggests that information retrieval via metadata be applied to various types of information. From this viewpoint, the technique of information retrieval via metadata could be regarded as the core technique. In addition, the author also discusses three levels of metadata: statically authority-controlled metadata, dynamically authority-controlled metadata, and user-oriented metadata.

隨著網際網路的日益普及，線上資源也越來越豐富，要精準的為讀者找出有用的資訊，前提是必須能夠精準的分析網頁內容。日期是網頁Metadata中的重要欄位，由於臺灣在日期格式的書寫習慣，使得中文網頁的日期形式較為複雜，因而增加了自動著錄網頁創造（或修改）日期時的困難。本研究的主要目的是針對網頁日期部分做深入的分析研究，以便能夠更精確的利用中文網頁中的日期欄位進行檢索利用。 本研究以隨機抽樣方式來抓取繁體中文網頁，分析及統計樣本網頁中出現的日期格式，並使用正規表示式來自動抓取正確的網頁日期，最後計算出正確率。透過此研究可以了解在進行中文網頁日期欄位自動辨識時可能會遭遇到的困難，並評估自動擷取繁體中文網頁日期欄位的可行性。實驗結果顯示，有日期資料網頁的正確率約為61%，沒有日期資料網頁的部分約為62%。有日期資料網頁的平均誤差年約為0.62年，且83.4%的網頁能精準預測其年份（即誤差年為0），因此雖然本研究的成果尚未能完全取代人工，但若應用得宜仍然可以提高網

Online resources have become more plentiful nowadays, thanks to the popularization of Internet services. In order to achieve accurate search results for the users, it is necessary to analyze web pages precisely. ‘Date’ is one of the most important fields of metadata in web pages. Due to the special date displaying formats using in Taiwan, it has made the automatic cataloging on date for webpage more difficult. The major purpose of this research is to thoroughly analyze different types of date displaying formats applied to Chinese web pages. These findings will be used to increase the precision on the date auto extraction of web pages. The procedures of experiment are as follows. Firstly, samples were randomly selected from Internet. Secondly, the statistic analysis on the date displaying format of each web pages was conducted. Lastly, Regular Expression was used to abstract the dates of each web page, while the accuracy ratio was also calculated. The difficulties and feasibility of auto date extraction are discussed in the end of this work. The results of the experiment suggest the accuracy ratio of web pages with date information is 61%. On the other hand, the accuracy ratio of web pages without date information is 62%. The average error of those web pages with date information is 0.62 year. The results of this research suggest that the auto date extraction mechanism can be used to improve the efficiency on webpage information retrieval.